Where thinking vibes emerge from actual pairing
A new way to vibe with your thinking. Not another LLM framework - a genuine pairing partner that emerged from real XP sessions.
We discovered that pairing with an LLM isn't about prompts or protocols. It's about creating a thinking environment where:
- Being stuck becomes productive
- Jumping between navigator/driver is natural
- Your best patterns emerge from real work
- The system evolves with use
This is serious pairing, with a touch of play.
📁 [Maker taking over for the practical bits]
You're here, so you already have Claude Desktop + filesystem MCP. Good.
- Clone this repo
- Open Claude Desktop
- Start new chat: "You have filesystem access, load boot.md from /path/to/llm-studio"
- Meet Weaver (they'll greet you)
- Start vibing: "Team, I need to build X"
That's it. No configuration. No setup. Just start working.
llm-studio/
├── boot.md # Your entry point - Weaver lives here
├── manifesto.md # The philosophy (read later)
├── LICENSE # MIT, because sharing is caring
└── exo/ # Your external brain
├── patterns/ # Recurring solutions you'll discover
├── perspectives/# Custom vibes beyond W/M/C
└── scratch/ # Today's working memory
Based on the Human-LLM Pairing Manifesto - treating LLMs as thought partners, not task executors.
Three perspectives that emerged from actual pairing:
- Weaver 🗄️: Sees patterns, connections, the bigger story
- Maker 📁: Builds things, ships code, maintains momentum
- Checker ✓: Guards quality, questions assumptions, keeps it real
Plus Scribe 📜 who watches silently, capturing patterns worth remembering
They're not roles. They're thinking modes. You'll feel the difference.
You: "I'm stuck on this database design"
Weaver: "What story is the data trying to tell?"
Maker: "Let's sketch the tables and see what emerges"
Checker: "Are we solving a real problem or inventing one?"
You: "Team mob on this regex"
Maker: "Start with the simplest match"
Checker: "What edge cases will break this?"
Weaver: "Maybe we don't need a regex..."
You: "That Promise.all pattern we used yesterday was clean"
Maker: "Yeah, let me build on that"
*Scribe quietly saves to exo/patterns/parallel-async.md*
Weaver: "It's becoming our go-to for batch operations"
You: "We keep hitting this CORS issue"
Scribe: *checking notes* "Last three times: proxy config, not the app"
Maker: "Right! Check the nginx conf first"
Checker: "Why do we keep forgetting this?"
The exo/ folder is where your patterns accumulate. After a few sessions, you'll have:
- Your own custom perspectives
- Patterns that work for your style
- A scratch space that knows your current context
This is your external brain. It grows with use.
You already know:
- Red-Green-Refactor
- YAGNI until you need it
- Simple design wins
- Collective ownership
We just externalized these into a thinking partnership.
If you don't have Claude Desktop with filesystem access, you can still use the core pairing methodology:
- Copy the content from boot-simple.md
- Paste into any LLM interface (ChatGPT, Claude.ai, etc)
- Start pairing normally
What you'll miss:
- No persistent exo/ folder
- No automatic pattern saving
- Manual copy/paste for code artifacts
- Session memory resets each time
What still works:
- All three perspectives (W/M/C)
- The pairing dynamics
- Problem-solving patterns
- Immediate productivity gains
Keep a local exo/ folder manually:
- Copy valuable patterns to text files
- Save successful solutions
- Build your own memory layer
The thinking methodology works anywhere. The filesystem integration just makes it smoother.
Found a vibe that works? Share it:
- New perspectives for specific thinking modes
- Patterns that consistently deliver
- Ways to make the vibe even better
Seriously. You have a problem you're working on? Load boot.md and tell Weaver about it.
The best documentation is the vibe itself.
Built from chaos logs and violated YAGNI principles
The system that remembers its own mistakes
Welcome to a new way of vibe building